项目文件夹

文件
wehub-resource-sync ba4be087d5
Create PR to main with cherry-pick from release / cherry-pick (push) Failing after 0s
CICD NeMo / pre-flight (push) Failing after 0s
CICD NeMo / configure (push) Has been skipped
Build, validate, and release Neural Modules / pre-flight (push) Failing after 1s
CICD NeMo / code-linting (push) Has been skipped
Build, validate, and release Neural Modules / release (push) Has been skipped
Build, validate, and release Neural Modules / release-summary (push) Has been cancelled
CICD NeMo / cicd-test-container-build (push) Has been cancelled
CICD NeMo / cicd-import-tests (push) Has been cancelled
CICD NeMo / L0_Setup_Test_Data_And_Models (push) Has been cancelled
CICD NeMo / cicd-main-unit-tests (push) Has been cancelled
CICD NeMo / cicd-main-speech (push) Has been cancelled
CICD NeMo / Nemo_CICD_Test (push) Has been cancelled
CICD NeMo / Coverage (e2e) (push) Has been cancelled
CICD NeMo / Coverage (unit-test) (push) Has been cancelled
CodeQL / Analyze (python) (push) Has been cancelled
CICD NeMo / cicd-wait-in-queue (push) Has been cancelled
chore: import upstream snapshot with attribution
2026-07-13 13:28:58 +08:00

67 行
2.1 KiB
Python

# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Implemented following the guide at https://www.promptingguide.ai/models/gemma#gemma-7b-prompt-format
"""
from lhotse.cut import Cut, MixedCut
from nemo.collections.common.data.prompt_fn import registered_prompt_format_fn
from nemo.collections.common.prompts.formatter import Modality, PromptFormatter
GEMMA_BOS = "<start_of_turn>"
GEMMA_END_OF_TURN = "<end_of_turn>"
GEMMA_NL = "\n\n"
class GemmaPromptFormatter(PromptFormatter):
NAME = "gemma"
OUTPUT_ROLE = "assistant"
INSERT_BOS = True
INSERT_EOS = True
TEMPLATE = {
"user": {
"template": f"{GEMMA_BOS}user\n|message|{GEMMA_END_OF_TURN}\n{GEMMA_BOS}model\n",
"slots": {
"message": Modality.Text,
},
},
OUTPUT_ROLE: {
# Note: that trailing NL is bothering me.
"template": f"|message|{GEMMA_END_OF_TURN}\n",
"slots": {
"message": Modality.Text,
},
},
}
@registered_prompt_format_fn(Cut, GemmaPromptFormatter)
def gemma1(cut: Cut, prompt: GemmaPromptFormatter):
if isinstance(cut, MixedCut):
cut = cut.first_non_padding_cut
if cut.has_custom("context"):
context = cut.context
elif cut.has_custom("question"):
context = cut.question
else:
context = cut.default_context
turns = [{"role": "user", "slots": {"message": context}}]
if (answer := cut.supervisions[0].text) is not None:
turns.append({"role": "assistant", "slots": {"message": answer}})
return prompt.encode_dialog(turns)